Method for determining hemodynamic parameters of a replacement heart valve

Abstract

The invention relates to a computer-implemented method for determining a hemodynamic parameter in a digital model of a heart of a subject, the method comprising receiving anatomical imaging data representing at least part of the heart; processing the anatomical imaging data to generate a patient-specific digital anatomical model; providing data describing a replacement heart valve; positioning a digital representation of the replacement heart valve within the digital anatomical model to form a composite model; determining patient-specific, time-varying motion fields of one or more anatomical structures of the heart; computing, using a physics-based blood-flow model applied to the composite model and driven by boundary conditions derived at least in part from the time-varying motion fields, a time-dependent blood-flow field; and determining, from the computed blood-flow field, at least one hemodynamic parameter characterising a physical property of blood flow through or around the replacement heart valve.

Description

[0001] METHOD FOR DETERMINING HEMODYNAMIC PARAMETERS OF A REPLACEMENT HEART VALVE

[0002] Technical field

[0003] The present invention relates to computer-implemented methods for analysing hemodynamic behaviour of replacement heart valves within patient-specific cardiac anatomy, based on medical imaging and computational blood-flow modelling.

[0004] Background

[0005] Heart-valve replacement procedures, including transcatheter aortic valve replacement (TAVR), transcatheter mitral valve replacement (TMVR), and surgical valve replacement (SAVR), are widely used to treat various forms of valvular heart disease. Bioprosthetic valves are commonly employed in these procedures, and their limited durability may necessitate re-intervention, such as valve-in-valve (ViV) implantation.

[0006] The hemodynamic performance of a replacement heart valve depends on numerous patient-specific factors, including cardiac anatomy, annular geometry, prosthesis size and orientation, and interactions between the valve and adjacent anatomical structures. Suboptimal valve selection, sizing, or positioning may contribute to complications such as paravalvular leakage, elevated transvalvular pressure gradients, coronary obstruction, patient-prosthesis mismatch, or adverse flow-jet impingement on surrounding tissue.

[0007] Pre-operative imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI), and echocardiography are routinely used to assess patient anatomy and to support planning of valve-replacement procedures. Computational modelling techniques have also been proposed for analysing aspects of cardiovascular anatomy or blood flow. However, existing approaches often rely on simplified or static anatomical representations, generic boundary conditions, or limited consideration of cardiac motion, which restricts their ability to represent patient-specific hemodynamic behaviour of replacement heart valves under realistic physiological conditions.

[0008] Accordingly, there remains a need for computational techniques that enable physiologically consistent, patient-specific evaluation of hemodynamic parameters associated with replacement heart valves, taking into account cardiac anatomy, valve geometry, and time-dependent blood-flow behaviour based on information obtainable from medical imaging and related clinical data. Summary

[0009] Heart-valve replacement procedures rely on pre-operative imaging to assess patient anatomy and to support planning of prosthetic valve implantation. The hemodynamic behaviour of a replacement heart valve is influenced by multiple patient-specific factors, including cardiac geometry, annular orientation, interactions with surrounding anatomical structures, and time-dependent blood-flow conditions associated with cardiac function. Existing analysis tools provide limited capability for evaluating valvespecific hemodynamic behaviour under realistic, patient-specific and time-varying anatomical and physiological conditions.

[0010] Accordingly, the present inventive concept provides computer-implemented techniques for determining hemodynamic parameters of a replacement heart valve positioned within a digital model of a subject’s heart. Anatomical imaging data are processed to generate a patient-specific anatomical model, and a digital representation of a replacement heart valve is positioned within the anatomical model to form a composite model representing interaction between the prosthesis and cardiac anatomy.

[0011] In some embodiments, time-varying motion of one or more cardiac structures is determined to capture dynamic deformation of the heart during at least a portion of cardiac function. The motion information may be obtained directly from imaging data or estimated using motion-estimation techniques, and is used to define dynamic boundary conditions or constraints for a physics-based blood-flow simulation. A time-dependent blood-flow field is thereby computed within the composite model, enabling determination of hemodynamic parameters associated with blood flow through or around the replacement heart valve.

[0012] The computational framework may be implemented using a variety of modelling approaches, including numerical fluid-dynamics solvers, reduced-order models, or machine-learning-based approximations. Anatomical structures and valve models may be represented using meshes, surface representations, implicit geometries, or other forms suitable for numerical simulation. The framework may further support evaluation of multiple valve configurations, including variations in valve type, size, implantation depth, rotational orientation, or valve-in-valve placement, to enable comparative analysis of resulting hemodynamic parameters.

[0013] The inventive concept thus provides a flexible and extensible platform for modelling hemodynamic behaviour of replacement heart valves using patient-specific anatomical, physiological, and motion-derived information obtainable from medical imaging and related clinical data. Preferred embodiments are set out in the dependent claims.

[0014] According to a first aspect of the present inventive concept, there is provided a computer-implemented method for determining a hemodynamic parameter in a digital model of a heart of a subject. The method comprises receiving input data that include anatomical imaging data representing at least part of the heart of the subject. The anatomical imaging data may be acquired as a single image dataset or as imaging data representing multiple cardiac phases.

[0015] Processing the imaging data comprises generating a digital anatomical model of the heart. In some embodiments, such processing comprises segmenting the imaging data to identify one or more anatomical structures, such as a ventricle, an atrium, an inflow tract, an outflow tract, an annulus region, or adjacent vessels. In other embodiments, the digital anatomical model is generated using model-based reconstruction, atlasbased modelling, machine-learning-based reconstruction, or other techniques that do not require explicit segmentation of the imaging data.

[0016] The digital anatomical model may represent static cardiac anatomy and / or may be associated with time-varying information describing motion or deformation of one or more cardiac structures. The digital anatomical model may be represented using one or more computer-readable forms, including surface representations, volumetric models, meshes, implicit geometries, or combinations thereof, suitable for computational analysis. The anatomical imaging data used to generate the digital anatomical model may be obtained using medical imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI), ultrasound imaging, or fluoroscopic imaging, and may be provided in one or more image datasets or image stacks, including DICOM data.

[0017] The method further comprises providing data describing a replacement heart valve. The valve data may include a digital representation of the prosthesis, such as a mesh, surface model, or implicit representation describing the geometry of a valve frame, valve leaflets, or other structural components. The valve data may additionally or alternatively include dimensional parameters, manufacturer-specified information, or other parameters that influence interaction between the valve and blood flow. As used herein, a replacement heart valve refers to a prosthetic valve configured to replace or supplement a native heart valve, including transcatheter or surgically implantable valves, and valve-in-valve configurations.

[0018] A digital representation of the replacement heart valve is positioned within the digital anatomical model to obtain a composite model. Positioning the valve may comprise aligning the valve model with an anatomical annulus, selecting an implantation depth or rotational orientation, or applying one or more rigid-body transformations defining a spatial configuration of the valve within the anatomy. In some embodiments, multiple candidate valve positions or orientations are evaluated. As used herein, a composite model refers to a combined representation of the digital anatomical model and the digital representation of the replacement heart valve, suitable for computational analysis.

[0019] The method further comprises determining patient-specific, time-varying motion fields of one or more anatomical structures of the heart. As used herein, time-varying motion fields refer to data describing temporal displacement, deformation, or motion of anatomical structures, represented, for example, as voxel-wise or mesh-based fields associated with the digital anatomical model.

[0020] In some embodiments, the anatomical imaging data comprise imaging data representing multiple cardiac phases, which may cover an entire cardiac cycle or only a portion thereof. When imaging data representing multiple cardiac phases are available, the motion fields may be determined using one or more motion-estimation techniques applied to the imaging data, including, for example, non-rigid image-registration techniques, feature-based tracking, optical-flow-based methods, deformable model fitting, or learning-based motion-estimation techniques.

[0021] In other embodiments, the anatomical imaging data do not directly provide time- resolved motion information over an entire cardiac cycle. For example, the imaging data may comprise a single cardiac phase or a limited subset of cardiac phases acquired due to scanner protocol, dose constraints, heart-rate variability, arrhythmia, or the clinical question. In such cases, time-varying motion of one or more anatomical structures may be estimated or inferred by completing missing motion phases using interpolation, extrapolation, periodic extension, and / or model-based inference techniques. The completion or inference of missing motion phases may be performed using one or more parametric motion models, learned motion models, atlas-based motion priors, biomechanical constraints, periodicity constraints, or combinations thereof. In some embodiments, auxiliary physiological measurements and / or complementary imaging modalities are incorporated to constrain or regularize the motion estimation, such that the resulting motion fields are physiologically consistent. Examples of such auxiliary information include stroke volume, ejection fraction, inflow or outflow waveforms, pressure estimates, volumetric changes of cardiac chambers, or flow measurements obtained from ultrasound or magnetic resonance imaging.

[0022] The inferred time-varying motion fields are associated with the digital anatomical model and are used to represent dynamic cardiac motion over time, even when the underlying imaging data do not directly capture the full temporal evolution of the heart.

[0023] Using the composite model, a physics-based blood-flow model is applied to compute a time-dependent blood-flow field within the anatomy. The blood-flow computation is driven by boundary conditions at least partially derived from the time-varying motion fields and may further incorporate physiological parameters of the subject. As used herein, boundary conditions define constraints applied to the blood-flow model, including moving-wall conditions derived from cardiac motion and inflow and / or outflow conditions. The flow model may comprise numerical solutions of fluid-dynamics equations, reduced-order modelling approaches, or machine-learning-based flow approximations, provided that the resulting blood-flow field varies over time.

[0024] A physics-based blood-flow model refers herein to a computational model that represents blood flow based on physical principles, such as conservation of mass and momentum, including numerical solvers, reduced-order models, or physics-informed machine-learning models.

[0025] Based on the computed blood-flow field, the method comprises determining one or more hemodynamic parameters. As used herein, hemodynamic parameters refer to quantities characterising physical properties of blood flow within the heart and through or around the replacement heart valve. Such parameters may include, without limitation, pressure gradients across the replacement heart valve, velocity or flow-rate distributions, flow-jet characteristics, vorticity, indicators of disturbed or turbulent flow, wall-shear-stress-related quantities, flow-impingement characteristics, residence time, or blood-stasis-related metrics. The hemodynamic parameters may be used to characterise blood-flow behaviour associated with the replacement heart valve and to assess the influence of valve type, size, implantation position, and / or orientation on the resulting flow field.

[0026] In some embodiments, the method enables interactive or automated evaluation of different valve configurations. The size, position, or type of the replacement valve may be selected by a user or determined automatically based on one or more computed hemodynamic parameters. The method may thereby support comparative assessment of multiple valve configurations within a computational environment.

[0027] The digital model of the heart may comprise the imaging data, the segmented anatomical model, the motion fields, and / or the computed blood-flow field, as appropriate. The method may be applied to data acquired using any suitable imaging technology and may be used for human or animal subjects.

[0028] In one example, the hemodynamic parameter is a flow parameter, the input data comprises a CT image of the heart of the subject, the processing of the input data comprises segmenting the CT image to identify components of the heart in the CT image, wherein the components of the heart comprise at least a ventricle and ascending aorta. Further, a fluid flow field may be determined based on the input data and the data parameters of the replacement heart valve when positioned in the digital model of the heart in at least one of the identified components of the heart. The method may further comprise determining, based on the fluid flow field, the flow parameter in the digital model of the heart.

[0029] In some embodiments, the replacement heart valve may be represented as a digital model of a prosthetic heart valve. The digital model may include information describing geometric, structural or functional characteristics of the physical valve. For example, the model may comprise three-dimensional data representing the valve frame, stent structure or leaflet surfaces, as well as parameters relating to valve type, manufacturer, nominal size or other specifications.

[0030] The digital model may also include data relating to the behaviour or motion of the prosthetic valve. Such data may represent, for instance, prescribed leaflet kinematics, deformation characteristics, rigid-body motion of the valve housing, or other functional or motional properties relevant to simulations of blood-flow interaction. These data may be used to simulate aspects of valve behaviour under different anatomical or physiological conditions.

[0031] In certain implementations, digital models of one or more prosthetic heart valves may be stored in a digital library, allowing a user or an automated system to select a valve model from a plurality of available options. The digital library may include models of different valve types, manufacturers or sizes, and the models may be categorised according to structural or functional parameters associated with the replacement heart valves. The selected digital model may be incorporated into the anatomical model of the heart to form the composite model used for the flow computation.

[0032] Using a digital representation of the prosthetic heart valve may allow different valve configurations to be evaluated computationally without requiring physical prototypes or surgical procedures. For example, multiple scenarios may be simulated efficiently in a digital environment by placing different valve models in different positions or orientations within the anatomical model. This may facilitate exploration of the behaviour of the prosthetic valve under varying anatomical, physiological or modelling conditions.

[0033] In some embodiments, the input data further comprise one or more physiological parameters associated with the subject and / or with a reference population. As used herein, physiological parameters refer to quantities that characterize blood properties, cardiac function, and / or boundary conditions of the cardiovascular system for use in the computational blood-flow model. Such physiological parameters may be patientspecific or population-derived, and may be obtained by direct measurement, estimated from medical imaging data, derived from clinical measurements, or assumed based on reference or normative values.

[0034] Physiological parameters may include, without limitation, blood density, blood viscosity (including parameters of Newtonian or non-Newtonian rheological models), hematocrit, temperature, heart rate, cardiac output, stroke volume, ejection fraction, atrial and / or ventricular pressures, systemic and / or pulmonary vascular resistance, vascular compliance, and parameters of lumped-parameter boundary condition models, such as Windkessel models. In some embodiments, the physiological parameters are used to define, constrain, or regularize boundary conditions and / or physical model parameters of the physics-based blood-flow simulation, thereby improving physiological consistency and patient specificity of the computed hemodynamic results. Variations in one or more physiological parameters may influence characteristics of the computed blood-flow field, including intra-cardiac flow patterns and the behaviour of blood flow through or around the replacement heart valve. In certain embodiments, the blood-flow field may be re-evaluated for different values of one or more physiological parameters to assess flow behaviour under different physiological conditions or to compare flow characteristics associated with different valve configurations.

[0035] In one embodiment of the present disclosure, the anatomical imaging data comprise computed tomography imaging, magnetic resonance imaging, ultrasound imaging, fluoroscopic imaging, or combinations thereof. The imaging modality may be selected based on clinical availability, the anatomical region of interest, spatial or temporal resolution requirements, or radiation-dose considerations. Computed tomography imaging may provide high spatial resolution suitable for defining cardiac chambers, annular geometry, and adjacent vessels, while magnetic resonance imaging may additionally provide soft-tissue contrast and functional information. Ultrasound or echocardiographic imaging may be used to supplement or constrain anatomical or functional information, for example by providing flow or velocity measurements, whereas fluoroscopic imaging may be employed to capture device positioning or motion during interventional procedures. The anatomical imaging data may be acquired using any suitable scanner configuration and protocol and may include contrast- enhanced or non-contrast acquisitions. The use of multiple imaging modalities may enable complementary information to be combined within the digital anatomical model, thereby improving robustness or fidelity of the subsequent computational analysis.

[0036] In one embodiment of the present disclosure, the anatomical imaging data comprise image frames corresponding to a plurality of cardiac phases, and the digital anatomical model is time-resolved across the plurality of cardiac phases. The plurality of cardiac phases may represent discrete time points distributed over an entire cardiac cycle or over a selected portion thereof, such as systole or diastole. Time-resolved anatomical modelling may be achieved by associating each image frame with a corresponding phase of the cardiac cycle and generating phase-specific anatomical representations, or by deriving continuous motion information between phases. The time-resolved digital anatomical model may thereby represent dynamic deformation, displacement, or motion of cardiac structures over time. Such representations may facilitate direct determination of motion fields from the imaging data and may reduce reliance on inferred or model-based motion estimation. The availability of time-resolved anatomy may improve correspondence between the simulated blood-flow field and the underlying physiological motion, and may enable more accurate evaluation of timedependent hemodynamic parameters.

[0037] In one embodiment of the present disclosure, the anatomical imaging data comprise imaging corresponding to a single cardiac phase or to a limited number of cardiac phases that do not fully resolve the temporal evolution of the heart over an entire cardiac cycle. Such imaging data may arise, for example, from single-phase acquisitions, prospectively gated scans, low-dose protocols, irregular heart rhythms, or clinical constraints that limit temporal sampling. In these cases, the digital anatomical model may be generated for the available cardiac phase or phases and may be augmented with time-varying information by estimating or inferring motion of one or more cardiac structures beyond the directly imaged phases. The estimation of cardiac motion may be performed using interpolation, extrapolation, periodic extension, or motion-completion techniques, and may rely on parametric motion models, learned motion representations, atlas-based priors, biomechanical constraints, or combinations thereof. Auxiliary physiological parameters or complementary imaging-derived measurements may be used to constrain or regularize the inferred motion such that the resulting time-varying motion fields are physiologically consistent. By enabling reconstruction of dynamic cardiac motion from incompletely time-resolved imaging data, the method may support time-dependent blood-flow simulation and hemodynamic analysis in situations where fully time-resolved imaging is unavailable.

[0038] In one embodiment of the present disclosure, the input data are received as DICOM data through a hospital information system or a picture archiving and communication system, implemented in a cloud-based or on-premise infrastructure. The reception of imaging data in DICOM format may enable standardized access to image pixel data, metadata describing acquisition parameters, and patient-related information, subject to applicable data- protection requirements. Integration with hospital information systems or PACS may allow automated or semi-automated retrieval of imaging data as part of a clinical workflow. The computational processing may be performed locally within a hospital network or remotely within a cloud-based computing environment, and data transfer may be secured using appropriate communication protocols. Such integration may facilitate deployment of the method in clinical or research settings and may support scalable computation or centralized data management. In one embodiment of the present disclosure, the input data further comprise one or more physiological parameters of the subject or of a reference population, and the physiological parameters are used to define or constrain boundary conditions and / or physical model parameters of the physics-based blood-flow model. Physiological parameters may represent quantities characterising blood properties, cardiac function, or vascular conditions, and may be obtained by direct measurement, derived from imaging data, inferred from clinical observations, or selected from reference values. By incorporating such parameters into the blood-flow model, the simulation may reflect subject-specific or physiologically plausible conditions, even when direct measurements are incomplete or unavailable. The physiological parameters may influence inflow or outflow conditions, moving-wall constraints, material properties, or other aspects of the computational model, and may thereby affect the resulting bloodflow field and computed hemodynamic parameters. In certain embodiments, the physiological parameters may be varied to evaluate sensitivity of the hemodynamic results to different physiological states or to compare simulated outcomes under alternative assumptions.

[0039] In one embodiment of the present disclosure, the digital anatomical model represents one or more anatomical structures of the heart. As used herein, anatomical structures refer to identifiable parts or regions of cardiac anatomy that may be resolved, modelled, or otherwise represented based on anatomical imaging data. In the present application, the terms “anatomical structures” and “components of the heart” may be used interchangeably to denote such identifiable parts of the cardiac anatomy, particularly in the context of processing imaging data to generate the digital anatomical model. The identified anatomical structures may comprise, without limitation, one or more of a ventricle, an atrium, an inflow tract, an outflow tract, a valve annulus, a left atrial appendage, heart valves, the ascending aorta, pulmonary veins, the inferior or superior vena cava, the pulmonary trunk, pulmonary arteries, or other anatomical regions represented in the anatomical model.

[0040] The particular set of anatomical structures that is identified may depend on the clinical context, the region of interest, and the quality or coverage of the anatomical imaging data. For example, ventricular and outflow tract structures may be identified for analysis of aortic valve replacement, whereas atrial structures or the left atrial appendage may be identified for mitral valve interventions or thromboembolic risk assessment. In some embodiments, the fluid-flow computation may further take into account effects associated with additional anatomical structures such as papillary muscles, chordae tendineae, or larger trabeculations. Such structures may be considered during image segmentation, during motion estimation using imageregistration or inference techniques, or during computation of the fluid-flow field itself. Incorporating these structures may provide increased geometric or motion detail in the anatomical representation, for example to improve local flow accuracy or interaction between blood flow and surrounding tissue.

[0041] In one embodiment of the present disclosure, processing the anatomical imaging data to generate the digital anatomical model comprises identifying one or more anatomical structures of the heart within the imaging data. The identification of anatomical structures may be implemented in various ways and may be performed using imagebased, model-based, or data-driven techniques. Such identification may comprise determining the presence, spatial location, extent, or topology of anatomical regions relevant to subsequent modelling steps, for example by delineating anatomical regions, extracting or classifying image content, assigning anatomical labels or parameters to image regions, or fitting one or more anatomical models to the imaging data.

[0042] In some embodiments, processing the anatomical imaging data comprises segmenting at least part of the anatomical imaging data, for example by applying one or more masks to delineate selected anatomical structures or regions of interest. Each mask may correspond to a particular anatomical structure or component, such as one or more of heart valves, ventricles, atria, the ascending aorta, pulmonary veins, the inferior or superior vena cava, the pulmonary trunk, pulmonary arteries, papillary muscles, or larger trabeculations of the heart, and a plurality of masks may be used to represent different regions or structures. In other embodiments, identification of anatomical structures may be assisted or performed using machine-learning methods, such as neural networks trained on annotated imaging data, or using rule-based, statistical, or atlas-based techniques. Such methods may be applied to imaging data representing a single cardiac phase, to time-resolved imaging data, or to imaging data acquired using different modalities. The identification of anatomical structures may therefore be achieved through segmentation or through other suitable computational approaches, without requiring a specific representation of all anatomical boundaries.

[0043] In one embodiment of the present disclosure, identifying one or more anatomical structures comprises extracting one or more anatomical structures from the imaging data or from intermediate representations derived therefrom. Extraction may include generating explicit surface or volumetric representations, implicit anatomical representations, or parametric descriptions associated with the identified anatomical structures. The extracted anatomical structures may be represented individually or jointly within the digital anatomical model and may serve as a basis for motion estimation, valve positioning, or flow simulation.

[0044] In one embodiment of the present disclosure, processing the anatomical imaging data to generate the digital anatomical model comprises applying a machine-learning model trained to identify, reconstruct, or refine cardiac anatomy from the anatomical imaging data. The machine-learning model may be configured to perform tasks such as anatomical detection, region labelling, shape reconstruction, refinement of anatomical boundaries, or completion of partially observed anatomy. The machine-learning model may operate directly on the imaging data or on derived features and may output explicit geometric representations, implicit anatomical fields, or parameters of a predefined anatomical model.

[0045] In one embodiment of the present disclosure, processing the anatomical imaging data further comprises generating a computational representation of the identified anatomical structures for use in the physics-based blood-flow model. The computational representation may comprise a volumetric mesh, a surface mesh, an implicit geometry, a level-set representation, or combinations thereof, and may be generated from segmented anatomical regions, reconstructed anatomical models, or directly from the imaging data. The representation may be adapted to the requirements of the selected blood-flow solver and may include refinement or smoothing in regions of interest, such as near valve leaflets or vessel walls.

[0046] In one embodiment of the present disclosure, the data describing the replacement heart valve comprise a digital valve model that includes geometric information representing one or more structural components of the prosthetic valve, such as a valve frame and / or one or more valve leaflets. The geometric information may describe overall valve shape, dimensions, and spatial relationships between valve components and may be suitable for positioning the valve within the digital anatomical model and for interaction with the blood-flow simulation. The digital valve model may represent the valve in a deployed or partially deployed configuration and may be adapted to different valve designs or implantation scenarios. In one embodiment of the present disclosure, the digital valve model further comprises information describing leaflet-motion behaviour. The leaflet motion may be defined by prescribed leaflet kinematics, for example by specifying time-dependent opening and closing trajectories, angular displacements, or deformation patterns associated with cardiac phases. In other embodiments, leaflet motion may be determined using a fluidstructure interaction model in which leaflet deformation is coupled to forces arising from the simulated blood flow and, optionally, from structural properties of the leaflet material. The choice between prescribed kinematics and fluid-structure interaction may depend on the desired balance between computational complexity and physical fidelity, and both approaches may be supported within the computational framework.

[0047] In one embodiment of the present disclosure, the data describing the replacement heart valve comprise manufacturer-specified dimensional parameters of a selected valve type or size. Such parameters may include nominal valve diameter, frame height, leaflet dimensions, or other specifications provided by the valve manufacturer. The manufacturer-specified parameters may be used to generate or select a corresponding digital valve model, to scale a generic valve representation, or to constrain valve placement within the digital anatomical model. Incorporating manufacturer-specified information may facilitate realistic representation of commercially available valve designs and may support evaluation of alternative valve sizes or models.

[0048] In one embodiment of the present disclosure, the data describing the replacement heart valve comprise valve parameters that influence at least one property of the simulated blood-flow field. Such valve parameters may affect flow resistance, jet formation, flow separation, or interaction between the blood flow and the valve structure. The valve parameters may be incorporated directly into the blood-flow model or may influence boundary conditions, material properties, or interaction models associated with the valve.

[0049] In one embodiment of the present disclosure, the valve parameters comprise one or more of frame stiffness, leaflet thickness, leaflet compliance, valve diameter, or deployment expansion characteristics. These parameters may be used, for example, to model deformation of the valve frame, flexibility of valve leaflets, or changes in valve geometry resulting from deployment within patient-specific anatomy. Variations in such parameters may influence the resulting blood-flow field and associated hemodynamic parameters and may be explored to assess sensitivity of valve performance to design or deployment characteristics.

[0050] In one embodiment of the present disclosure, the digital valve model is stored or represented as a three-dimensional mesh, a surface model, or an implicit surface representation. The chosen representation may be compatible with the selected bloodflow solver and may enable efficient numerical simulation of flow through and around the valve. The digital valve model may be generated from manufacturer data, from imaging-derived reconstructions, from parametric templates, or from combinations thereof, and may be converted between different representations as needed for positioning, motion modelling, or flow computation.

[0051] In one embodiment of the present disclosure, positioning the digital representation of the replacement heart valve comprises aligning a geometric annulus of the valve model with an anatomical annulus of the digital anatomical model. The anatomical annulus may be identified based on imaging data, reconstructed anatomical geometry, or derived anatomical landmarks, and may represent a native valve annulus or an annular region associated with a previously implanted prosthesis. Alignment may involve matching size, orientation, or spatial location of the geometric annulus of the valve model with the anatomical annulus, and may be performed automatically, semi- automatically, or manually, for example using optimization techniques or user-guided adjustment. Such alignment may improve anatomical correspondence between the valve model and the surrounding cardiac anatomy.

[0052] In one embodiment of the present disclosure, positioning the digital representation of the replacement heart valve comprises orienting the valve model along a longitudinal centerline of an inflow tract or outflow tract of the anatomical model. The centerline may be derived from the digital anatomical model using geometric analysis, skeletonization, or model-based fitting techniques. Orienting the valve model relative to the centerline may facilitate physiologically consistent alignment with the direction of blood flow and may reduce geometric interference with surrounding anatomical structures. The orientation may be adjusted to account for anatomical curvature, asymmetry, or patient-specific variations.

[0053] In one embodiment of the present disclosure, positioning the digital representation of the replacement heart valve comprises selecting a target depth or implantation level relative to an anatomical annulus or to one or more adjacent cardiac structures. The implantation depth may be defined as a distance along the inflow or outflow direction, a relative position with respect to anatomical landmarks, or a fraction of valve height extending into adjacent chambers or vessels. The selected depth may influence interaction between the valve and surrounding anatomy and may affect flow patterns, pressure gradients, or other hemodynamic parameters. The implantation depth may be selected based on clinical planning considerations, anatomical constraints, or computational evaluation of alternative positions.

[0054] In one embodiment of the present disclosure, the method comprises positioning the digital representation of the replacement heart valve at a plurality of candidate positions within the digital anatomical model and computing a corresponding plurality of timedependent blood-flow fields. The candidate positions may differ in one or more of alignment, orientation, depth, or rotational configuration. The corresponding hemodynamic parameters derived from the simulated blood-flow fields may be compared to assess the influence of valve positioning on flow behaviour. In some embodiments, one of the plurality of candidate positions may be identified based on a comparison of the corresponding hemodynamic parameters, for example to identify a position associated with reduced pressure gradients, improved flow symmetry, or reduced flow stasis.

[0055] In one embodiment of the present disclosure, positioning the digital representation of the replacement heart valve comprises applying a rigid-body transformation that defines a three-dimensional spatial configuration of the valve model within the digital anatomical model. The rigid-body transformation may include one or more translations and rotations applied to the valve model relative to the anatomical model. The transformation may be determined based on anatomical landmarks, alignment criteria, optimization objectives, or user input. In some embodiments, the rigid-body transformation may be combined with scaling or deformation operations, for example to account for deployment expansion or interaction with patient-specific anatomy.

[0056] In one embodiment of the present disclosure, the anatomical imaging data comprise a single cardiac phase or a subset of cardiac phases that does not fully resolve cardiac motion over an entire cardiac cycle, and the time-varying motion fields are inferred to represent cardiac motion over time. The subset of cardiac phases may correspond to a limited number of image frames acquired at discrete time points or may represent partial temporal coverage of the cardiac cycle due to clinical, technical, or physiological constraints. In such embodiments, cardiac motion over time may be reconstructed by inferring motion between and beyond the imaged phases, thereby enabling representation of dynamic cardiac deformation despite incomplete temporal imaging data.

[0057] In one embodiment of the present disclosure, the inferred or directly determined timevarying motion fields define moving-wall boundary conditions or a deforming computational domain for the physics-based blood-flow model. The motion fields may be used to prescribe time-dependent displacement or deformation of anatomical boundaries, such as chamber walls or vessel walls, and may thereby influence local and global blood-flow patterns within the computational domain. Representing cardiac motion through moving boundaries or deforming domains may enable the blood-flow simulation to reflect physiologically relevant interactions between cardiac motion and intravascular flow.

[0058] In one embodiment of the present disclosure, determining the time-varying motion fields comprises applying a non-rigid image registration technique to anatomical imaging data representing multiple imaging phases. The non-rigid image registration may estimate spatial correspondences between image frames acquired at different cardiac phases and may generate displacement or deformation fields describing motion of anatomical structures over time. Such registration techniques may be applied pairwise between successive phases or jointly across multiple phases and may be configured to preserve anatomical continuity or topology.

[0059] In one embodiment of the present disclosure, the non-rigid image registration technique comprises a free-form deformation model, a diffeomorphic registration model, or an optical-flow-based model. These models may represent motion using control-point grids, smooth invertible transformations, or local displacement vectors, respectively, and may be selected based on desired trade-offs between computational efficiency, robustness, and anatomical fidelity. The registration technique may further incorporate regularization terms or constraints to promote smoothness or physiological plausibility of the estimated motion fields.

[0060] In one embodiment of the present disclosure, determining the time-varying motion fields comprises estimating cardiac motion for one or more non-imaged phases using interpolation, extrapolation, biomechanical or atlas-based motion models, machinelearning models, and / or physiological measurements used to constrain the motion estimation. The estimation of motion for non-imaged phases may rely on assumed periodicity of cardiac motion, learned motion patterns from reference populations, biomechanical relationships between pressure, volume, and deformation, or statistical correlations captured by trained models. Physiological measurements, such as stroke volume, ejection fraction, chamber volume changes, or flow waveforms, may be incorporated to regularize or constrain the inferred motion, thereby improving physiological consistency of the resulting motion fields.

[0061] In one embodiment of the present disclosure, the time-varying motion fields comprise a voxel-wise or mesh-based displacement field associated with the digital anatomical model. The displacement field may describe spatial motion of anatomical structures at discrete time points or continuously over time and may be defined on image voxels, mesh nodes, surface elements, or other computational representations. Such motion fields may be directly coupled to the blood-flow solver or may be interpolated or mapped between different representations used for anatomical modelling and flow simulation.

[0062] In one embodiment of the present disclosure, determining the time-dependent bloodflow field comprises combining motion information associated with the replacement heart valve with motion information of one or more anatomical structures of the heart. Valve-motion information, such as leaflet kinematics or deformation of a valve frame, and cardiac-motion information derived from the time-varying motion fields may be incorporated into the computational model such that the simulated blood flow reflects relative motion between the replacement heart valve and surrounding anatomical structures. In some embodiments, the blood-flow computation may further incorporate pressure conditions or flow conditions associated with anatomical regions located upstream or downstream of the replacement heart valve. By combining valve-related motion, anatomical motion, and physiological boundary conditions, the computed blood-flow field may represent dynamic interactions between the replacement heart valve and the cardiac anatomy under time-varying conditions.

[0063] In some embodiments, the effect of cardiac motion on the blood-flow simulation is represented, at least in part, through prescribed boundary conditions in the form of a time-varying flow-rate profile and / or pressure waveform. For example, a flow-rate profile at an inflow and / or outflow boundary may be derived from the determined timevarying motion fields (e.g., based on time-varying chamber volume change implied by the motion fields) and used to drive the physics-based blood-flow model. In further embodiments, such flow-rate profiles and / or pressure waveforms are additionally or alternatively constrained using one or more physiological parameters, such as heart rate, stroke volume, ejection fraction, inflow / outflow measurements, or pressure estimates, to obtain physiologically consistent time-dependent boundary conditions, particularly in cases where the available imaging data comprise a single cardiac phase or a limited subset of cardiac phases.

[0064] In one embodiment of the present disclosure, computing the time-dependent blood-flow field comprises solving governing equations of fluid motion within the composite model representing the digital anatomical model and the positioned replacement heart valve. In some embodiments, the governing equations comprise the Navier-Stokes equations for incompressible or weakly compressible flow, which may be solved numerically to obtain velocity, pressure, or related flow quantities as functions of space and time. The equations may be discretised using finite-volume, finite-element, finite-difference, lattice-Boltzmann, or other suitable numerical methods, and may be solved over a deforming computational domain defined by time-varying anatomical motion and valve behaviour.

[0065] In one embodiment of the present disclosure, computing the time-dependent blood-flow field comprises modelling interaction between valve leaflets and the blood flow using a fluid-structure interaction model. The fluid-structure interaction model may couple the blood-flow solver with a structural or kinematic model of the valve leaflets and, in some embodiments, the valve frame. Such coupling may account for forces exerted by the blood flow on the leaflets and resulting leaflet deformation or motion, and may be implemented using monolithic or partitioned solution schemes. In alternative embodiments, leaflet motion may be prescribed or parameterised while still allowing interaction with the surrounding flow field to be represented in an approximate or reduced-order manner.

[0066] In one embodiment of the present disclosure, computing the time-dependent blood-flow field comprises using a physics-informed neural network trained to approximate a solution of a blood-flow model governed by physical constraints. The physics-informed neural network may be trained to satisfy governing equations of fluid motion, such as conservation of mass and momentum, together with boundary conditions and, where applicable, initial conditions. Such neural-network-based approaches may be used alone or in combination with numerical solvers, for example to accelerate computation, to provide surrogate models for repeated evaluation of valve configurations, or to approximate flow solutions in regions of interest. In some embodiments, one or more steps of the method, including flow computation, motion estimation, or boundarycondition inference, may be supported or carried out by neural networks trained on manually curated ground-truth data or simulated reference data.

[0067] In one embodiment of the present disclosure, the boundary conditions used in the physics-based blood-flow model comprise patient-specific physiological waveforms and / or physiological parameter values. The boundary conditions may include, for example, time-varying inflow or outflow waveforms, pressure conditions, or impedancebased boundary models associated with anatomical regions upstream or downstream of the replacement heart valve. Such boundary conditions may be derived from anatomical imaging data, from image-registration-based motion analysis, from complementary imaging modalities providing flow or motion information, or from correlations between physiological quantities such as flow rates, chamber volumes, or cardiac muscle mass. Incorporating patient-specific physiological waveforms or parameter values may improve correspondence between the simulated blood-flow field and subject-specific cardiac function.

[0068] In one embodiment of the present disclosure, the fluid-flow field may thus be determined based on a combination of anatomical imaging data, motion information of cardiac structures and the replacement heart valve, physiological boundary conditions, and computational fluid-dynamics algorithms. By integrating these sources of information, the computed blood-flow field may represent dynamic interactions between blood flow, cardiac anatomy, and valve behaviour under time-varying physiological conditions, thereby enabling subsequent determination of hemodynamic parameters associated with the replacement heart valve.

[0069] In some embodiments, the time-dependent blood-flow field or one or more boundary conditions used in the physics-based blood-flow model may be determined or prescribed based on correlations between physiological quantities of the heart and flow characteristics. For example, an inflow or outflow condition may be estimated based on a relationship between cardiac muscle mass, chamber volume, stroke volume, peak flow rate, or combinations thereof, or may be selected from a predefined look-up table derived from reference or population data. Such flow information may be constant or time-varying and may be patient-specific or population-derived. In some implementations, a prescribed flow field or prescribed boundary conditions may be used instead of, or in combination with, a numerically computed blood-flow field.

[0070] In one embodiment of the present disclosure, one or more steps of the method may be supported or carried out using one or more neural networks. The neural networks may be trained on manually curated ground-truth data, on synthetically generated training data, on simulation-derived data, or on combinations thereof, depending on the respective task. The neural networks may comprise convolutional neural networks, recurrent neural networks, graph-based neural networks, transformer-based architectures, or other suitable machine-learning models, and may be trained in a supervised, semi-supervised, or unsupervised manner.

[0071] In some embodiments, a neural network may be configured to process the anatomical imaging data to identify or reconstruct anatomical structures of the heart. Such processing may include, for example, segmentation, anatomical labelling, shape reconstruction, refinement of anatomical boundaries, or completion of partially observed anatomy from imaging data acquired using one or more modalities. In further embodiments, a neural network may be configured to provide data associated with the replacement heart valve, such as parameters describing valve geometry, deployment characteristics, leaflet behaviour, or other functional properties relevant to interaction with blood flow.

[0072] In some embodiments, a neural network may be configured to support determination of the time-varying motion fields of cardiac structures, for example by estimating motion from time-resolved imaging data, by inferring motion from sparse or single-phase imaging data, or by learning motion patterns from reference populations. Such neural- network-based motion estimation may be used alone or in combination with imageregistration techniques, biomechanical models, or physiological constraints.

[0073] In some embodiments, a neural network may be configured to support or perform computation of the blood-flow field based on the input data and the data describing the replacement heart valve positioned within the digital anatomical model. For example, a neural network may approximate solutions of governing equations of fluid motion, may act as a reduced-order or surrogate model for a physics-based solver, or may provide corrections or accelerations to numerically computed flow fields. In further embodiments, a neural network may be configured to determine one or more hemodynamic parameters based on a computed or estimated blood-flow field, including, for example, pressure gradients, velocity distributions, flow-jet characteristics, or indicators of wall impingement or flow obstruction.

[0074] In one embodiment of the present disclosure, any of the neural-network-based components described herein may be used in combination with, or as an alternative to, algorithmic, physics-based, or rule-based processing steps. Neural networks may thereby support or supplement one or more steps of the method by providing data estimates, classifications, parameter predictions, or other outputs relevant to the computational framework, while allowing flexibility in modelling fidelity and computational performance.

[0075] In some embodiments, the fluid-flow field may be estimated or simulated using numerical, reduced-order, data-driven, or hybrid computational approaches, depending on the desired modelling fidelity, available computational resources, and characteristics of the input data.

[0076] In one embodiment of the present disclosure, the hemodynamic parameter comprises a pressure gradient across the replacement heart valve. The pressure gradient may be determined based on the computed time-dependent blood-flow field by evaluating pressure values upstream and downstream of the replacement heart valve. In some embodiments, the pressure gradient may be quantified as a mean pressure difference over at least a portion of a cardiac cycle or over a period during which the replacement heart valve is open. In other embodiments, the pressure gradient may be expressed as a peak pressure difference, a peak-to-peak pressure difference, or another representative measure characterising pressure loss across the valve. Such pressuregradient measures may be used to assess valve performance, flow resistance, or patient-prosthesis interaction.

[0077] In one embodiment of the present disclosure, the hemodynamic parameter comprises one or more flow-related quantities, such as velocity magnitude, flow rate, vorticity, or turbulent kinetic energy. These quantities may be derived from the computed bloodflow field and may characterise flow behaviour within cardiac chambers, through the replacement heart valve, or in adjacent vessels. In some embodiments, flow velocity or flow rate may be evaluated locally or globally, for example at one or more crosssections or regions of interest. Vorticity or turbulence-related measures may be used to identify disturbed flow patterns, flow separation, or regions of increased energy dissipation.

[0078] In one embodiment of the present disclosure, the hemodynamic parameter may further comprise an energy-loss or viscous-dissipation measure derived from the computed blood-flow field. Such a measure may characterise loss of kinetic energy due to viscous effects within the fluid and may be evaluated over a defined anatomical region or over time. The energy-loss measure may be computed, for example, based on fluid viscosity and local rates of deformation of the flow field and may be integrated spatially and / or temporally to obtain a representative value.

[0079] In one embodiment of the present disclosure, the hemodynamic parameter comprises a wall shear stress or an oscillatory shear index on one or more anatomical or prosthetic surfaces. The wall shear stress may be computed on surfaces representing cardiac anatomy, such as chamber walls or vessel walls, and / or on surfaces representing components of the replacement heart valve, such as the valve frame or leaflets. The oscillatory shear index may be derived from temporal variations of the wall shear stress and may provide information on directional changes of shear forces over time. Such parameters may be relevant for assessing potential tissue response, thrombogenic risk, or durability of prosthetic components.

[0080] In one embodiment of the present disclosure, the hemodynamic parameter comprises a blood-residence-time metric or a blood-stasis metric within a cardiac chamber or in proximity to the replacement heart valve. The blood-residence-time metric may characterise the duration that blood particles remain within a defined region, while blood-stasis metrics may identify regions of low flow velocity or recirculation. Such parameters may be evaluated, for example, within an atrium, within a ventricle, or adjacent to prosthetic valve structures, and may be used to assess conditions associated with thrombus formation or impaired washout.

[0081] In one embodiment of the present disclosure, determining the hemodynamic parameter comprises computing one or more time-averaged quantities over at least a portion of a cardiac cycle. Time-averaged quantities may include mean velocity, mean pressure gradient, time-averaged wall shear stress, or other averaged measures derived from the time-dependent blood-flow field. In some embodiments, the averaging interval may correspond to a full cardiac cycle, to a systolic or diastolic phase, or to a period during which the replacement heart valve is open or closed. Time-averaged and peak-based measures may be used individually or in combination to characterise hemodynamic behaviour associated with the replacement heart valve.

[0082] In a further aspect of the present disclosure, there is provided a computer system comprising one or more processors and a memory storing instructions which, when executed by the one or more processors, cause the computer system to perform the method according to any of claims 1-37. The computer system may be configured to receive anatomical imaging data, physiological data, and data describing a replacement heart valve, and to process such data to generate digital anatomical models, determine time-varying motion fields, compute time-dependent blood-flow fields, and determine one or more hemodynamic parameters as described herein. The computer system may further be configured to output computed results, intermediate data, or visualisations associated with the method.

[0083] In some embodiments, the computer system may be implemented as a processing unit comprising one or more processors configured to execute instructions, one or more memory units for storing data and executable instructions, and one or more input and output interfaces for data exchange. The computer system may be implemented in a cloud-based environment, an on-premise computing environment, or a distributed computing architecture combining local and remote resources. In some embodiments, the computer system may be configured to receive imaging data from a user device, from a hospital information system or picture archiving and communication system, or directly from one or more imaging modalities, such as a computed tomography scanner, magnetic resonance imaging system, or ultrasound device. The computer system may further be configured to access or store digital representations of replacement heart valves, anatomical models, motion fields, or other data used in connection with the method.

[0084] In a further aspect of the present disclosure, there is provided a computer program product comprising instructions stored on a non-transitory computer-readable medium which, when executed by one or more processors, cause the one or more processors to perform the method according to any of claims 1-37. The computer-readable medium may comprise any suitable non-transitory medium, including but not limited to semiconductor memory, magnetic storage, optical storage, or combinations thereof. The computer program product may be distributed, stored, or executed in any suitable computing environment, including cloud-based, on-premise, or hybrid environments. The computer system and computer program product according to these further aspects may incorporate any of the features, embodiments, or variations described in relation to the method aspects of the present disclosure. Features and advantages corresponding to those of the method may likewise apply to the computer system and computer program product, and combinations of features described in relation to different aspects are contemplated unless explicitly stated otherwise.

[0085] In one embodiment of the present disclosure, the method may further comprise providing an interactive visualisation of the digital anatomical model of the heart. The interactive visualisation may present the digital anatomical model and, where applicable, the positioned digital representation of the replacement heart valve on a display or other graphical interface. The interface may allow a user to interact with the model, for example by selecting, positioning, orienting, or configuring a digital representation of a replacement heart valve within the anatomical model. The timedependent blood-flow field may then be determined based on the selected position, orientation, or configuration of the replacement heart valve.

[0086] In some embodiments, the interactive visualisation may enable exploration of different valve configurations, including variations in valve type, size, orientation, or manufacturer-specific models, for example by selecting valve representations from a digital library. The valve may be positioned at one or more anatomical locations represented in the digital anatomical model, such as annular regions associated with aortic, mitral, tricuspid, or pulmonary valves, and multiple valve configurations may be evaluated sequentially or in parallel.

[0087] The interactive visualisation may include graphical tools such as zooming, rotation, transparency control, or highlighting of selected anatomical regions or valve components. In some embodiments, the interactive visualisation may additionally display one or more hemodynamic parameters derived from the computed blood-flow field, including, for example, pressure gradients, flow velocities, flow-jet characteristics, wall shear stress distributions, or indicators of flow impingement or obstruction. Time- resolved visualisations of simulated blood-flow behaviour may also be presented.

[0088] In certain embodiments, the interactive visualisation may be provided together with a functional report comprising numerical, graphical, or tabulated representations of hemodynamic or functional parameters derived from the simulation, such as flow rates, pressures, volumes or volume changes of cardiac structures, or pressure differences across the replacement heart valve. A user interacting with the visualisation may include technical operators, analysts, researchers, or other users engaging with computational representations of cardiac anatomy and simulated flow behaviour.

[0089] In the present disclosure, references to “one embodiment” or “some embodiments” refer to illustrative examples and do not limit the scope of the disclosure to a single embodiment. Features described in relation to one embodiment may be combined with features described in relation to other embodiments unless such a combination is technically incompatible.

[0090] Where the present disclosure refers to a component, element, or step being configured to perform a function, such references are intended to encompass any suitable implementation, including algorithmic, model-based, data-driven, or hardware- supported implementations, unless explicitly stated otherwise.

[0091] The methods described herein may be implemented in any suitable order, and the order of method steps is not intended to be limiting unless a specific order is explicitly required. In some embodiments, certain steps may be performed in parallel, repeated, omitted, or combined, depending on the available input data or the intended application.

[0092] The features described in relation to the method aspects of the present disclosure may be implemented correspondingly in a computer system or a computer program product as described herein. Likewise, features described in relation to the computer system or computer program product may be applied to the method aspects, unless explicitly stated otherwise.

[0093] The use of the terms “comprise”, “comprises”, or “comprising” does not exclude the presence of elements or steps other than those listed. The use of the terms “a”, “an”, or “the” in reference to an element does not exclude a plurality of such elements.

[0094] Reference signs, if any, included in the claims or description are provided for ease of understanding and do not limit the scope of the claims.

[0095] Description of the drawings

[0096] The above, as well as additional objects, features, and advantages of the present inventive concept, will be better understood through the following illustrative and non- limiting detailed description, with reference to the appended drawings. In the drawings like reference numerals will be used for like elements unless stated otherwise.

[0097] Fig. 1 is a block diagram of a method for determining a hemodynamic parameter.

[0098] Fig. 2 is a block diagram of a method for determining a flow parameter.

[0099] Fig. 3 schematically illustrates a cross-section of a human heart.

[0100] Fig. 4 schematically illustrates a method for determining flow parameters in a digital model of a heart of a subject.

[0101] Figs. 5A-C illustrate a digital model of a heart comprising a replacement heart valve, including the interactive positioning and an exemplary replacement heart valve with leaflets.

[0102] Detailed description

[0103] Fig. 1 illustrates a block diagram of a method for determining a hemodynamic parameter in a digital model of a heart of a subject.

[0104] The method (1000) may comprise the steps of:

[0105] • (1100) receiving input data comprising multi-phase anatomical imaging data of at least part of the heart of the subject;

[0106] • (1200) processing the input data to generate a digital anatomical model of the heart, for example by identifying one or more anatomical components represented in the imaging data;

[0107] • (1300) providing data describing a replacement heart valve;

[0108] • (1400) positioning a digital representation of the replacement heart valve within the anatomical model to form a composite model;

[0109] • (1500) determining time-varying motion fields of one or more anatomical structures of the heart by applying a non-rigid image-registration algorithm to the multi-phase anatomical imaging data;

[0110] • (1600) computing a time-dependent blood-flow field within the composite model using a physics-based blood-flow model driven by boundary conditions derived from the time-varying motion fields; and • (1700) determining from the computed blood-flow field a hemodynamic parameter characterising a physical property of blood flow through the replacement heart valve.

[0111] The method illustrated in Fig. 1 represents a generic embodiment. The steps may be implemented using any suitable data-processing techniques, computational models, or imaging modalities, and the order of steps may be varied in accordance with alternative embodiments described herein.

[0112] Fig. 2 illustrates a block diagram of an example embodiment of the method, in which computed tomography (CT) imaging is used as the anatomical input data.

[0113] The method (2000) may comprise:

[0114] • (2100) providing input data comprising a CT image of the heart of the subject;

[0115] • (2200) processing the CT image to identify components of the heart, such as a ventricle or the ascending aorta; the processing may include segmenting the CT image or applying one or more masks or machine-learning techniques;

[0116] • (2300) providing data describing a replacement heart valve;

[0117] • (2400) determining a fluid-flow field based on the CT-derived anatomical model and the data describing the replacement heart valve, for example when the valve is positioned in at least one of the identified components of the heart; and

[0118] • (2500) determining, based on the computed fluid-flow field, one or more flow- related parameters, which may include, for example, a pressure difference across the replacement heart valve, a flow velocity through the valve, a jetdirection measure, or parameters associated with wall impingement or outflow obstruction.

[0119] In certain embodiments, the fluid-flow field of step (2400) may be computed using a computational fluid-dynamics (CFD) algorithm or another numerical flow-modelling technique. One or more steps of the method may be supported or implemented by a neural network trained on ground-truth data, for example for segmenting the CT image or estimating flow-related quantities.

[0120] Although not illustrated in Fig. 2, additional steps may include outputting values of the determined flow parameters or visualising the computed fluid-flow field or anatomical model. Fig. 3 illustrates a schematic cross-sectional representation of a human heart (10).

[0121] The figure depicts several anatomical structures that may be identified from the imaging data or represented in a digital anatomical model. As shown, the heart (10) may comprise a left atrium (LA), left ventricle (LV), ascending aorta (Aao), mitral valve (MV), aortic valve (AV), right ventricle (RV), right atrium (RA), tricuspid valve (TV) and pulmonary valve (PV). Additional structures may include the pulmonary veins, the inferior and superior vena cava, the pulmonary trunk and the pulmonary arteries.

[0122] It will be understood that the heart may comprise further anatomical components not illustrated in Fig. 3, and that the level of anatomical detail represented in a digital model may vary depending on the available imaging data and the intended computational analysis.

[0123] In some embodiments, the anatomical structures illustrated in Fig. 3 may be derived from processing anatomical imaging data, such as CT, MRI, ultrasound or other modalities. The identification of these structures may involve segmentation, masking, classification or other computational procedures suitable for extracting anatomical information from the imaging data.

[0124] Fig. 4 schematically illustrates an example of a system for carrying out a method, for example the method illustrated in Fig. 1, for determining a hemodynamic parameter in a digital model (410) of a heart (10) of a subject (100).

[0125] Anatomical imaging data of the heart (10) may be acquired from a medical imaging device, such as a computed tomography (CT) scanner, magnetic resonance imaging (MRI) scanner, ultrasound device, fluoroscopic imaging system, or another medical imaging modality. The imaging data may be provided as input data (200) and may include one or more images (210) representing at least a portion of the heart (10). The anatomical imaging data may comprise a single image dataset or imaging data representing multiple cardiac phases. The input data (200) may be supplied directly from an imaging device, from a local storage system, or via a hospital information system (HIS), which may comprise or interface with a picture archiving and communication system (PACS). The imaging data may be provided in a standard medical-data format, such as DICOM, although other formats may be used. In some embodiments, the input data (200) may additionally comprise physiological information related to the subject (100), such as heart rate, pressure-related values, cardiac output, or other physiological parameters suitable for use with the computational framework. Such physiological information may be used to define or constrain boundary conditions or model parameters of a blood-flow simulation and / or to support estimation of cardiac motion.

[0126] The input data (200) may be received at a processing environment, which may be implemented using a cloud-based infrastructure (300), an on-premises infrastructure, or a combination thereof. The processing environment may execute instructions that perform one or more steps of the method described herein, including processing the imaging data to generate a digital anatomical model (410), providing and positioning a digital representation of a replacement heart valve (420) within the anatomical model to form a composite model, determining patient-specific time-varying motion fields of one or more anatomical structures, computing a time-dependent blood-flow field using a physics-based blood-flow model driven by boundary conditions derived at least in part from the motion fields, and determining one or more hemodynamic parameters from the computed flow field.

[0127] In some embodiments, the time-varying motion fields are determined from imaging data representing multiple cardiac phases using one or more motion-estimation techniques. In other embodiments, when time-resolved motion information is not directly available from the imaging data, the motion fields are estimated or inferred based on the available imaging data, optionally in combination with physiological parameters, biomechanical models, statistical motion models, or learned models.

[0128] In the example of Fig. 4, the processing environment is further connected to a server (500), which may facilitate data management, communication, storage, or distribution of results. The server (500) may provide data to one or more visualisation devices (400), such as a workstation or display, which may present images of the digital anatomical model (410), the positioned replacement heart valve (420), and / or representations of the computed blood-flow field. The visualisation device (400) may additionally present one or more derived hemodynamic parameters or other outputs produced by the method.

[0129] In some embodiments, the processing is performed using a computer program product comprising instructions stored on a non-transitory computer-readable medium and executed by one or more processors. Input data and output data may be communicated through a PACS, HIS, or other clinical information systems, although this is not required.

[0130] The output of the method may include one or more values derived from the computed time-dependent blood-flow field, such as pressure gradients across the replacement heart valve (420), velocity or flow-rate measures, flow-jet characteristics, wall-shear- stress-related quantities, residence time, or indicators of flow impingement or outflow obstruction. Such results may be presented on the visualisation device (400) and / or stored for further analysis or comparison across multiple valve configurations.

[0131] Fig. 5 illustrates examples of visualisations of a digital anatomical model (410) of a heart (10) and a digital representation of a replacement heart valve (420).

[0132] These visualisations may be generated as part of the method described herein, for example after processing anatomical imaging data to derive a digital model of the heart and positioning the replacement heart valve (420) within that model.

[0133] In the example of Fig. 5(A), the digital model (410) depicts anatomical structures of the heart (10), and the replacement heart valve (420) is shown positioned within or between one or more components of the heart. This visualisation may be static or time- resolved, depending on the available imaging data and the computational model used.

[0134] Fig. 5(B) illustrates that the replacement heart valve (420) may be displayed in association with a user interface enabling adjustment of one or more parameters of the digital representation of the valve, such as its position, orientation or other characteristics. In some embodiments, such adjustments may be performed interactively by a user operating a visualisation device (400). Any changes made to the placement or configuration of the replacement heart valve (420) may be applied to the composite model and may be used in determining the corresponding blood-flow field and derived flow parameters.

[0135] Fig. 5(C) illustrates, by way of example, different representations (422) of the replacement heart valve (420), for instance showing a closed and an open configuration, or different geometric or functional states. Such representations may be based on valve-specific data, including structural or motion-related information associated with the corresponding physical valve design. A visualisation device (400) may present the digital anatomical model (410), the positioned replacement heart valve (420) and, in some embodiments, output derived from the computed time-dependent blood-flow field, such as hemodynamic parameters. The visualisation device (400) may be any suitable display or interactive computing system, including a workstation, tablet, monitor or other interface configured to receive and present data associated with the method.

[0136] Examples

[0137] The following descriptions provide illustrative examples of how the inventive concept may be implemented. These examples are non-limiting and are provided solely to demonstrate various ways in which the method, systems and models described herein may be realised. The inventive concept is not restricted to any of the particular imaging modalities, anatomical-processing techniques, computational models, numerical algorithms, parameter selections, or user-interaction mechanisms described in these examples.

[0138] Different combinations of features may be used, and alternative implementations may be employed without departing from the scope of the claims. The examples below should therefore be understood as optional embodiments that complement, but do not limit, the general description of the invention.

[0139] Example 1: Exemplary Workflow for Determining a Fluid-Flow Field

[0140] The following describes an exemplary methodology that may be used for determining a time-dependent fluid-flow field in a digital anatomical model of the heart.

[0141] This example illustrates one possible implementation; other workflows, imaging modalities, algorithms and parameter choices may be used.

[0142] Anatomical imaging data, such as CT, MRI, ultrasound or other imaging modalities, may be processed to identify cardiac structures. For illustration, structures such as a ventricle and the ascending aorta — examples of which are shown schematically in Fig. 3, may be extracted at one or more phases of the cardiac cycle. The resulting anatomical model may be exported in a 3D file format (e.g., STL) or otherwise represented for computational use. Additional anatomical components may be included where relevant for flow modelling. The extracted anatomical structures may be subject to optional surface regularisation and re-meshing, for example to obtain a surface representation compatible with the chosen computational method. Mesh density, element type and meshing strategy may be selected as appropriate for the modelling framework.

[0143] When multi-phase or time-resolved imaging is available, time-varying motion of the anatomical structures may be determined using one or more motion-estimation techniques. In one example, a non-rigid image-registration algorithm is applied to successive image phases to produce a displacement field describing motion of points on the cardiac structures over time. In other examples, motion may be determined using feature-based tracking, optical-flow-based methods, deformable model fitting, or learning-based motion-estimation techniques applied to the imaging data. Interpolation techniques, such as cubic Hermite interpolation, may optionally be used to obtain a temporally continuous motion field from discrete imaging phases. In further embodiments, motion information may be derived or constrained using complementary imaging modalities and / or physiological correlations, such as volume change, inflow or outflow measurements, or pressure-related parameters.

[0144] The anatomical model, replacement heart valve model and any available motion fields may then be used as input to a physics-based blood-flow model. Various numerical approaches may be employed, including but not limited to computational fluid dynamics (CFD) solvers, immersed-boundary methods, finite-element or finite-volume formulations, reduced-order models, physics-informed neural networks, or hybrid numerical and data-driven models. As an example, an immersed-boundary approach may represent the fluid on an Eulerian mesh and the anatomical structures on a Lagrangian mesh with optional adaptive refinement. Boundary conditions may be defined using pressure, flow or lumped-parameter models, such as variants of Windkessel models or alternative physiological inflow / outflow conditions. Solver settings such as time-stepping, numerical schemes and fluid parameters may be chosen according to implementation requirements, and alternate choices may be used.

[0145] A digital representation of a replacement heart valve, illustrated schematically in Fig. 5, may include geometric, structural or motion-related information. Valve motion in a model may include rigid-body components associated with cardiac motion and leaflet dynamics influenced by pressure or flow. As an example, rigid-body motion of the valve frame may be estimated by identifying valve-tissue contact regions (“landing zone”) in the anatomical model and applying an alignment algorithm such as Horn’s quaternionbased method. Other approaches may be used for estimating valve motion or incorporating leaflet behaviour.

[0146] In some embodiments, a user may interact with a digital anatomical model, as illustrated in Fig. 5A and Fig. 5B, to position, orient or select the replacement heart valve within the model. These user-defined configurations may be incorporated into the computational procedure used to determine the corresponding blood-flow field. Fig. 5C illustrates an example of multiple valve configurations or leaflet states that may be represented.

[0147] This, and other, exemplary workflow is provided for illustrative purposes only. Numerous variations are possible, including different imaging modalities, anatomical- processing techniques, modelling fidelities, valve-representation methods and computational algorithms. All such variations remain within the scope of the inventive concept as defined in the appended claims.

[0148] Example 2: Non-Limiting Workflow Using MRI, ML-Based Segmentation, and Reduced- Order Flow Modelling

[0149] A further example of a workflow for determining a time-dependent blood-flow field is described below. This example uses magnetic resonance imaging (MRI) as the anatomical input data and demonstrates that the inventive concept is not limited to CT- based or CFD-based implementations.

[0150] Anatomical MRI data of the heart may be obtained as a series of multi-phase images capturing the motion of cardiac structures over the cardiac cycle. Automated processing of the MRI data may be performed using a machine-learning model trained to identify relevant anatomical regions. The resulting anatomical representation may include the ventricles, atria and adjacent vessels, although the level of detail may vary depending on the MRI sequence and spatial resolution.

[0151] Time-varying motion information may be derived directly from the multi-phase MRI images. For example, a neural-network-based motion-estimation model may be used to extract point-wise displacements between successive phases, or a conventional non-rigid registration method may be applied. The estimated motion fields may be smoothed or interpolated in time to obtain a continuous representation of cardiac dynamics.

[0152] The replacement heart valve may be represented by a digital model that includes geometric information and valve-specific motion behaviour. As an example, the opening and closing dynamics of the valve leaflets may be approximated using a reduced-order model trained from bench-top measurements or manufacturer-provided data. The rigid-body component of valve motion may be inferred from the cardiac motion field, for instance by determining a best-fit transformation between the valveseating region and the corresponding region in the anatomical model.

[0153] A reduced-order blood-flow model may be used to compute the time-dependent flow field in the composite model formed by the heart anatomy and the positioned replacement heart valve. For example, a physics-informed neural network may be trained to satisfy the governing equations of blood flow subject to boundary conditions derived from the cardiac motion and inflow / outflow measurements. Alternatively, a lumped-parameter network model may be used to estimate pressure and flow distributions within regions of interest. Other modelling approaches may also be employed.

[0154] On the basis of the computed flow field, one or more hemodynamic parameters may be determined. These parameters may be derived either directly from the model or by post-processing of the estimated pressure or velocity fields. Examples include pressure gradients across the replacement heart valve, flow-jet direction, local shear metrics or indicators of flow stasis.

[0155] A user interface may optionally be provided to allow interactive adjustment of the position or orientation of the replacement heart valve within the digital anatomical model, and the resulting changes in the estimated flow field may be displayed.

[0156] Example 3: Echocardiography + Atlas-Based Anatomy + Hybrid ML / Physics Flow Solver

[0157] A further non-limiting example illustrates how the inventive concept may be implemented using echocardiographic imaging and atlas-based anatomical inference. This example demonstrates that the method is not limited to CT or MRI data and can be applied in situations where imaging data offers lower spatial coverage or temporal resolution.

[0158] In this example, two- or three-dimensional echocardiographic images may be acquired over multiple cardiac phases. While such data may provide only a partial view of the cardiac anatomy, anatomical structures of interest may nevertheless be extracted using machine-learning models trained for ultrasound image segmentation. The extracted structures may include the left ventricle, outflow tract or regions surrounding the intended valve-implantation site. A statistical shape model or anatomical atlas may then be used to infer missing anatomical regions or to complete the geometric representation of the heart. The completed anatomical model may be generated in a format suitable for computational analysis.

[0159] Time-varying motion of the anatomical structures may be obtained from the multiphase echocardiographic images. In one embodiment, cardiac motion may be estimated using a learning-based optical-flow method adapted for ultrasound, optionally combined with a non-rigid registration step to increase temporal consistency. The resulting motion field may be projected onto the inferred anatomical surfaces to generate a time-varying anatomical model compatible with subsequent flow modelling.

[0160] A digital representation of a replacement heart valve may be incorporated into this anatomical model. In this example, the valve may be represented using a simplified geometric model combined with a learned surrogate model describing leaflet dynamics. The surrogate model may be trained using bench measurements, manufacturer- provided motion data or numerical simulations of the valve under physiological conditions.

[0161] The composite model may be used as input to a hybrid flow-modelling framework. For example, a physics-informed neural network may be trained to approximate velocity and pressure fields that satisfy, in a weak sense, the governing equations of fluid motion as well as boundary conditions derived from the cardiac motion field. The network may be regularised using prior knowledge of valvular flow patterns or by incorporating predictions from a reduced-order fluid model. Such hybrid approaches may allow flow predictions even in cases where full-resolution CFD is impractical due to limited imaging data or computation time. From the estimated flow field, one or more hemodynamic parameters may be derived. These may include a surrogate of trans-valvular pressure drop, estimates of flow acceleration through the prosthetic valve, or indicators of flow recirculation in the ascending aorta or outflow tract. These values may be displayed on a visualisation device together with the anatomical model and the digital representation of the valve.

[0162] As in other examples, a user interface may optionally allow a user to reposition, rotate or replace the digital representation of the replacement heart valve. The hybrid flow model may be re-evaluated for each configuration, enabling rapid comparison of different implantation scenarios.

[0163] Example 4: Single-Phase Imaging with Inferred Cardiac Motion

[0164] A further non-limiting example illustrates an implementation of the inventive concept in which time-varying anatomical motion is estimated using single-phase imaging data.

[0165] In this example, anatomical imaging data are obtained as a single cardiac phase, such as an end-diastolic computed tomography (CT) image acguired during routine preoperative planning. The imaging data are processed to generate a digital anatomical model of the heart, including anatomical structures relevant to replacement valve implantation, such as the ventricle, annulus region, and adjacent vessels. The anatomical model may be generated using segmentation, atlas-based reconstruction, machine-learning-based modelling, or combinations thereof.

[0166] Because the imaging data do not directly provide time-resolved information describing cardiac motion, time-varying motion fields of one or more anatomical structures are estimated using motion-estimation techniques. In one embodiment, a parametric or biomechanical cardiac motion model is employed, and model parameters are adapted based on the available anatomical geometry and one or more physiological parameters of the subject, such as heart rate, stroke volume, or ejection fraction. In other embodiments, a learned motion model trained on population data is used to infer subject-specific cardiac motion from the single-phase anatomical input.

[0167] A digital representation of a replacement heart valve is positioned within the digital anatomical model to form a composite model. The valve may be aligned with an anatomical annulus and placed at a selected implantation depth or orientation. The inferred time-varying motion fields are associated with the anatomical model and used to define dynamic boundary conditions for a physics-based blood-flow simulation.

[0168] Using the composite model, a time-dependent blood-flow field is computed using a physics-based blood-flow model. Boundary conditions may be derived from the inferred motion fields and may further incorporate physiological parameters or lumped- parameter boundary models. Based on the computed blood-flow field, one or more hemodynamic parameters associated with blood flow through or around the replacement heart valve are determined, such as pressure gradients, flow-rate distributions, or indicators of flow stasis.

[0169] Items

[0170] 1. A computer-implemented method (1000) for determining a flow parameter in a digital model (410) of a heart (10) of a subject (100), the method (1000) comprising:

[0171] • providing (1100) input data (200) comprising a computed tomography, CT, image (210) of the heart (10) of the subject (100),

[0172] • segmenting (1200) the CT image (210) to identify components of the heart (10) in the CT image (210), wherein the components of the heart (10) comprise at least a ventricle and ascending aorta,

[0173] • providing (1300) data parameters of a replacement heart valve (420),

[0174] • determining (1400) a fluid flow field based on the input data (200) and the data parameters of the replacement heart valve (420) when positioned in the digital model (410) of the heart (10) in at least one of the identified components of the heart (10), and

[0175] • based on the fluid flow field, determining (1500) the flow parameter in the digital model (410) of the heart (10).

[0176] 2. The computer-implemented method according to item 1, wherein determining the fluid flow field further comprises combining motion data of the replacement heart valve in the digital model of the heart with motion data of the heart.

[0177] 3. The computer-implemented method according to item 1 or 2, wherein the flow parameter comprises a pressure difference over the replacement heart valve, wherein the pressure over the replacement heart valve is quantified by a mean pressure difference over a period when the replacement heart valve is open, a peak pressure difference, and / or a peak-to-peak pressure difference between a peak pressure upstream of the replacement heart valve and a peak pressure downstream of the replacement heart valve. The computer-implemented method according to any one of the preceding items, wherein the fluid flow field is further determined based on the CT image, utilizing image registration on the segmented components of the heart, or by using flow or motion information from an imaging modality or by using a correlation between flow rates and cardiac muscle mass of the heart. The computer-implemented method according to any one of the preceding items, wherein the input data further comprises a physiological parameter related to the heart of the subject, such as: heart rate, blood pressure, myocardial contractibility, estimation of aortic and lung venous pressure, wherein the fluid flow field is further determined based on the physiological parameter. The computer-implemented method according to any one of the preceding items, wherein the method further comprises providing an interactive visualization of the heart of the subject, wherein an interface for the interactive visualization of the heart is configured to enable a user to position the replacement heart valve in the interactive visualization of the heart, and wherein the fluid flow field is determined based on a placement of the replacement heart valve by the user in the interactive visualization of the heart. The computer-implemented method according to any one of the preceding items, wherein the replacement heart valve is a digital model of a prosthetic heart valve. The computer-implemented method according to any one of the preceding items, wherein the input data is provided in form of medical standard data such as DICOM data, the input data being received via a Hospital Information System, HIS, in a cloud-based and / or on-premise infrastructure, wherein the HIS comprises a Picture Archiving and Communication System, PACS. The computer-implemented method according to any one of the preceding items, wherein one or more of the steps of the method is supported and / or carried out by one or more neural networks, wherein the neural networks are trained on manually curated ground truth data.

[0178] 10. The computer-implemented method according to any one of the preceding items further comprising, outputting a pressure difference value based on the determined flow field, a flow velocity value based on the flow field through the replacement heart valve, a flow jet direction, a flow jet wall impingement, outflow obstructions, viscous energy loss, and / or fluid shear stress.

[0179] 11. The computer-implemented method according to any one of the preceding items, wherein the components of the heart comprise a heart valve, ascending aorta, ventricles, atriums, pulmonary veins, inferior and superior vena cava, pulmonary trunc and arteries, and / or an effect of papillary muscles and / or larger trabeculations of the heart.

[0180] 12. The computer-implemented method according to any one of the preceding items, wherein the segmenting of the CT image comprises applying a mask on the CT image to identify the components of the heart.

[0181] 13. The computer-implemented method according to any one of the preceding items, wherein determining the fluid flow field comprises applying a computational fluid dynamics algorithm.

[0182] 14. A non-transitory computer-readable medium storing instructions thereon which, when executed by a processor, cause the processor to carry out the steps of the method (1000) according to item 1.

[0183] 15. A processing unit configured to receive input data (200) comprising a computed tomography, CT, image (210) of a heart (10) of a subject (100), wherein the processing unit is configure to execute the steps of: a. segmenting the CT image (210) to identify components of the heart (10) in the CT image (210), wherein the components of the heart (10) comprises at least a ventricle and ascending aorta, b. providing data parameters of a replacement heart valve (420), c. determining a fluid flow field based on the input data (200) and the data parameters of the replacement heart valve (420) when positioned in a digital model (410) of the heart (10) in at least one of the identified components of the heart (10), and d. based on the fluid flow field, determining a flow parameter in the digital model (410) of the heart (10).

Claims

Claims1. A computer-implemented method for determining a hemodynamic parameter in a digital model of a heart of a subject, the method comprising: a) receiving input data comprising anatomical imaging data representing at least a part of the heart of the subject; b) processing the input data to generate a digital anatomical model of the heart; c) providing data describing a replacement heart valve; d) positioning a digital representation of the replacement heart valve within the digital anatomical model to form a composite model; and e) determining patient-specific, time-varying motion fields of one or more anatomical structures of the heart; f) computing, using a physics-based blood-flow model applied to the composite model and driven by boundary conditions derived from the timevarying motion fields, a time-dependent blood-flow field in the composite model; and g) determining, from the computed blood-flow field, a hemodynamic parameter characterising a physical property of blood flow through the replacement heart valve.

2. The method of claim 1, wherein the anatomical imaging data comprise computed tomography (CT) imaging, magnetic resonance imaging (MRI), ultrasound imaging, or fluoroscopic imaging.

3. The method of any of the preceding claims, wherein processing the anatomical imaging data to generate the digital anatomical model comprises identifying the one or more anatomical structures.

4. The method of claim 3, wherein the identified anatomical structures comprise a ventricle, an atrium, an inflow tract, an outflow tract, a valve annulus, and / or a left atrial appendage.

5. The method of any of the preceding claims, wherein processing the anatomical imaging data to generate the digital anatomical model comprises segmenting at least part of the anatomical imaging data.

6. The method of claim 5, wherein segmenting the anatomical imaging data comprises extracting one or more of the anatomical structures.

7. The method of any of the preceding claims, wherein processing the anatomical imaging data to generate the digital anatomical model comprises applying a machine-learning model trained to reconstruct or refine cardiac anatomy from the anatomical imaging data..

8. The method of any of the preceding claims, wherein the anatomical imaging data comprise image frames corresponding to a plurality of cardiac phases, and the digital anatomical model is time-resolved across the plurality of cardiac phases.

9. The method of any of the preceding claims, wherein processing the anatomical imaging data further comprises generating a computational mesh or surface representation of the anatomical structures for use in the physics-based bloodflow model.

10. The method of any of the preceding claims, wherein the input anatomical imaging data are received as DICOM data through a hospital information system or PACS in a cloud-based or on-premise infrastructure.11 . The method of any of the preceding claims, wherein the input data further comprise one or more physiological parameters of the subject or of a reference population, and wherein the physiological parameters are used to define or constrain boundary conditions and / or physical model parameters of the physicsbased blood-flow model.

12. The method of any of the preceding claims, wherein the data describing the replacement heart valve comprise a digital valve model including geometry of a prosthetic valve frame and / or valve leaflets.

13. The method according to claim 12, wherein the digital valve model further comprises leaflet-motion behaviour defined by prescribed leaflet kinematics or by a fluid-structure interaction model.

14. The method of any of the preceding claims, wherein the data describing the replacement heart valve comprise manufacturer-specified dimensional parameters of a selected valve type or size.

15. The method of any of the preceding claims, wherein the data describing the replacement heart valve comprise valve parameters that influence at least one property of the simulated blood-flow field.

16. The method according to claim 15, wherein the valve parameters comprise one or more of: frame stiffness, leaflet thickness, leaflet compliance, valve diameter, or deployment expansion characteristics.

17. The method of any of the preceding claims, wherein the digital valve model is stored as a three-dimensional mesh, surface model, or implicit surface representation.

18. The method of any of the preceding claims, wherein positioning the digital representation of the replacement heart valve comprises aligning a geometric annulus of the valve model with an anatomical annulus of the digital anatomical model.

19. The method of any of the preceding claims, wherein positioning the digital representation of the replacement heart valve comprises orienting the valve model along a longitudinal centerline of an inflow tract or outflow tract of the anatomical model.

20. The method of any of the preceding claims, wherein positioning the digital representation of the replacement heart valve comprises selecting a target depth or implantation level relative to an anatomical annulus or adjacent cardiac structure.

21. The method of any of the preceding claims, wherein the method comprises positioning the digital representation of the replacement heart valve at a plurality of candidate positions and computing a corresponding plurality of timedependent blood-flow fields, and optionally identifying one of the plurality of candidate positions based on a comparison of the corresponding hemodynamic parameters.

22. The method of any of the preceding claims, wherein positioning the digital representation of the replacement heart valve comprises applying a rigid-body transformation defining a three-dimensional spatial configuration of the valve model within the anatomical model.

23. The method of any of the preceding claims, wherein the anatomical imaging data comprise a single cardiac phase or a subset of cardiac phases, and the time-varying motion fields are inferred to represent cardiac motion over time.

24. The method of any of the preceding claims, wherein the time-varying motion fields define moving-wall boundary conditions or a deforming computational domain for the physics-based blood-flow model.

25. The method of any of the preceding claims, wherein determining the timevarying motion fields comprises applying a non-rigid image registration technique to anatomical imaging data representing multiple imaging phases.

26. The method according to claim 25, wherein the non-rigid image registration technique comprises a free-form deformation model, a diffeomorphic registration model, or an optical-flow-based model.

27. The method of any of the preceding claims, wherein determining the timevarying motion fields comprises estimating cardiac motion for one or more nonimaged phases using interpolation, extrapolation, biomechanical or atlas-based motion models, machine-learning models, and / or physiological measurements used to constrain the motion estimation.

28. The method of any of the preceding claims, wherein the time-varying motion fields comprise a voxel-wise or mesh-based displacement field associated with the digital anatomical model.

29. The method of any of the preceding claims, wherein computing the timedependent blood-flow field comprises solving the Navier-Stokes equations for blood flow within the composite model.

30. The method of any of the preceding claims, wherein computing the timedependent blood-flow field comprises modelling interaction between the valve leaflets and the blood flow using a fluid-structure interaction model.

31. The method of any of the preceding claims, wherein computing the timedependent blood-flow field comprises using a physics-informed neural network trained to approximate a solution of a blood-flow model governed by physical constraints.

32. The method of any of the preceding claims, wherein the boundary conditions comprise a prescribed time-varying flow-rate profile and / or pressure waveform derived from the time-varying motion fields and / or constrained using one or more physiological parameters of the subject.

33. The method of any of the preceding claims, wherein the boundary conditions used in the physics-based blood-flow model comprise patient-specific physiological waveforms and / or physiological parameter values.

34. The method of any of the preceding claims, wherein the hemodynamic parameter comprises a pressure gradient across the replacement heart valve.

35. The method of any of the preceding claims, wherein the hemodynamic parameter comprises one or more of: velocity magnitude, flow rate, vorticity, or turbulent kinetic energy.

36. The method of any of the preceding claims, wherein the hemodynamic parameter comprises a wall shear stress or oscillatory shear index on one or more anatomical or prosthetic surfaces.

37. The method of any of the preceding claims, wherein the hemodynamic parameter comprises a blood-residence-time metric or a blood-stasis metric within a cardiac chamber, such as an atrium or in proximity to the replacement heart valve.

38. The method of any of the preceding claims, wherein determining the hemodynamic parameter comprises computing one or more time-averaged quantities over at least a portion of a cardiac cycle.

39. A computer system comprising one or more processors and memory storing instructions which, when executed by the one or more processors, cause the computer system to perform the method of any of claims 1-38.

40. A computer program product comprising instructions stored on a non-transitory computer-readable medium which, when executed by one or more processors, cause the one or more processors to perform the method of any of claims 1-38.

Images